IBM's future: 6 things it needs to thrive

IBM has now paid Globalfoundires $1.5B to take over all of its semiconductor manufacturing operations, which used to be known as IBM Microelectronics.

As a result Big Blue is now "Fabless", or lacking its own chip fabrication capability, like its enterprise competitors HP, Fujitsu and Oracle, that also sell big iron enterprise servers that run on UNIX, but have processors that are produced by others.

HP uses Intel for its Itanium-based Nonstop/Superdome systems, and Oracle uses TSMC to build its UltraSPARCs, which is GlobalFoundries's primary rival. Fujitsu partners with LSI Corp for its own SPARC-architecture chips.

GlobalFoundries is a California company that was formed after AMD divested its own chip manufacturing capacity and merged with Chartered Semiconductor Manufacturing. It is owned by the royal family of the Persian Gulf nation of Abu Dhabi, which is part of the United Arab Emirates (UAE). 

Among chips that IBM produced as a semiconductor manufacturer for the embedded systems industry (such as the PowerPC that is the CPU in the XBOX 360, PS3 and Nintendo consoles, as well as in other industry applications) the East Fishkill, New York and Essex Junction, Vermont plants that GlobalFoundries is taking over produce the POWER architecture chips that go into IBM's System p UNIX/Linux boxes as well as their System z mainframes. 

In addition to their PureSystems integrated appliances, these are the only computer systems IBM still manufactures since they sold off their Intel-based server line to Lenovo, as it did with their PC business in 2004.

As we know from Apple, which is also a "fabless" company, not having native chipmaking capability isn't necessarily a bad thing. But we know from history that brain drain and also losing fundamental expertise in chip design could be disastrous.

HP also was once a company that could fabricate its own chips, as did Digital Equipment corporation which it had acquired via the Compaq merger, which produced the Alpha supercomputing chip. After being purchased by Compaq, DEC killed off the Alpha, and sold the manufacturing capabilities to Samsung (which eventually, had the Alpha's bus technologies licensed to AMD and eventually, ended up at GlobalFoundries).

Later on HP decided to end its PA-RISC processor line, terminate its own production capabilities, and went into partnership with Intel on the Itanium.  

And we all know how that one turned out. 

If the System p and the System z is to continue to be IBM's crown jewels and best in breed big iron that enterprises have come to rely on, then IBM must stay actively involved in the engineering process of the silicon.

While GlobalFoundries has agreed to supply IBM with chips for 10 years and also provide employment for those engineers, there is always the risk of brain drain. GlobalFoundries may, at some point, decide that maintaining operations in the United States is not cost effective.

For that reason alone, IBM must maintain a core set of superior engineering talent for its POWER chips in-house or risk repeating HP's past mistakes.

In addition to the POWER systems themselves and maintaining technical superiority in the enteprise server space, the other big crown jewel that IBM has is Watson, its natural language processing and data analytics platform that was made famous back in February of 2011 by beating champions Ken Jennings and Brad Rutter in a highly publicised, no-holds-barred Jeopardy! match.

Since then IBM has made inroads with Watson in select industry verticals, such as in medical diagnostics and health insurance, by selling Watson as an integrated hardware and software platform. However because of its dependency (or rather its optimization) on the System p's derivative of Linux and its large scale supercomputing requirements in terms of infrastructure needed to deploy it, few enterprises have had the wherewithal to stand up a Watson installation on-premises.

Because of the huge CAPEX required to stand up a Watson, obviously the next logical step is to bring Watson to the cloud, as a pure subscriber service. IBM has literally just begun to do this in October of 2014 with Watson Services for BlueMix.

This is a beta public cloud offering that is specifically tailored for developing apps that run on Watson services, which includes Concept Expansion, Language Identification, Machine Translation, Message Resonance, Question & Answer, Relationship Extraction and User Modelling.

Today, Watson Services are for developer types. But what enterprises and vertical businesses want are usable apps. Ideally, IBM will set up an online marketplace/app store and developer program and a community for this in order to produce what Watson apps businesses actually want.

While there is no question the technology is exciting and powerful, Watson Services for BlueMix is very much geeky stuff with no apps to show for yet. SoftLayer's Watson for Analytics being the only exception, and it is unclear whether it actually uses BlueMix or not.

There's also the issue that many businesses and verticals -- particularly those with data soverignity issues -- may not be comfortable with their data co-existing with other customers on IBM's public cloud. They may want a private cloud implementation.

In addtition to running Watson workloads on BlueMix and its own in-house provider, Softlayer, IBM needs to form partnerships with large-scale services providers and specialized vertical ISVs that are willing to put up Watson apps as private SaaS. And while the POWER architecture is unmatched in its class for this type of application, IBM may wish to consider porting Watson to x86, for smaller private implementations.

Considering that IBM has already entered a technology-sharing agreement with Intel, maybe Watson's future might not be closed-system and proprietary after all.  

In addition to Watson in the cloud, there are traditional workloads that IBM needs to capitalize on using its own datacenters.

IBM has been making strategic cloud investments for a while, such as with its 2013 purchase of Softlayer. However Softlayer has been operating very much in a silo and has been left, more or less to function independently without direct interference from its parent company.

As with Watson, IBM Big Iron servers, namely the System p and System z products are expensive to deploy, and even the largest of enterprises are now considering whether or not it makes sense to keep those workloads running in-house or to find a way to strategically outsource them.

Although IBM Global Technology Services is finding it difficult to maintain a pipeline signing new datacenter consolidation engagements -- which has become increasingly evident in the past several quarters -- it makes sense for IBM to treat its own datacenters -- and its own public cloud as consolidation targets it can use for strategic outsourcing purposes.

After all, only IBM has the potential to offer under one roof the ability to cloud migrate on-premises IBM-centric workloads that currently run on AIX, Mainframe and Intel also Intel. And to do it in both a public and private cloud model all in-house.

The big question is whether or not it can execute in a seamless, self-service and rapidly provisioned fashion that its competitors currently do with Windows and Linux-based IaaS, such as at AWS and Azure, or if it is going to be more of a very high-touch, managed hosting type of thing that requires a great deal of (expensive) IBM outsourced sysadmin babysitting, as it does already.

The IBM/Apple partnership, which was announced back in July , has re-kindled both interest in Big Blue's software and also potentially  the iPad and iPhone in the enterprise.

Still, other than some initial lip service when the partnership was launched, and a single slide during this month's iPad Air 2 launch event that quoted an IBM Vice President praising Apple's Swift development language, not much has transpired from the partnership as of yet.

While there is no question that Apple's iOS represents a very large opportunity for IBM to make headroads in the mobile space, it would be foolish of Big Blue to exclusively tie itself to a single platform, regardless of whatever favored nation status it gives to Cupertino.

If IBM is in fact dedicating thousands of developers to the partnership in order to write mobile apps for iOS, then allocating some of their time towards producing Android and modern Windows apps also makes sense as well. After all, despite IBM's efforts on the software and cloud services side, the iPhone and iPad are hardly enterprise-class devices.

Doing new and innovative things is clearly important as part of the recipe for IBM's comeback. Unfortunately, so is downsizing.

IBM is perhaps the largest employer in the entire IT industry, to the tune of 430,000+ employees serving in over 170 countries, of which approximately 190,000 are engaged in services alone.

As you can see from the above slide, services is where IBM is currently hurting the most.  And services income, in addition to software, is where IBM is supposed to be making up the slack from the Systems & Technology (hardware) division, which has now been largely divested as a result of  the Lenovo and Globalfoundries transactions.

As part of a $1B "workforce rebalancing" that started in June of 2013, IBM already went through a large round of layoffs, to the tune of anywhere between 8,000 and 10,000 people globally. But that may be just a drop in the bucket compared to what might actually be needed to stop the bloodletting and to make Wall Street happy again.

IBM may need as much as a 20 percent or even higher headcount reduction over the next few years, which would be the largest it has had to do since the early 1990s when the company almost foundered entirely. But for the same reasons low growth numbers upset investors, "Resource Actions" on that scale may be very detrimental to the company's reputation as well as affect the agility of its sales force, and could also hurt its market valuation. 

One of the ways that IBM may be able to silently rid itself of unwanted headcount is to incentivize people to leave. Early retirements for staff in their 50's may be in the cards -- while at the same time keeping some of these folks on as contractors who still have valauable expertise. Not having to pay benefits has its advantages.

Additionally, rolling out salary reductions for employees needing "new training" may just bring about a voluntary exodus closer to the numbers they want without actually having to do the deed. 

Sure, it's demoralizing, but when you are dealing in numbers this big, there are few elegant solutions.

Virginia Rommety, who was promoted from within and took the reigns as IBM's CEO in late 2012 has arguably become one of the most divisive in the company's history.

While IBM was initially applauded by finally giving a woman the helm of one of the world's most envied technology companies, within the company itself, Rometty is feared by many and morale is now at an all-time low, according to many current and former IBMers who I have spoken with.

But don't just take my word for it. I think Steve Denning at Forbes nailed it when he outlined the core of IBM's problems as being largely leadership-related:

• A relentless (arguably reckless) pursuit of cost cutting
• Shifting towards cheaper expertise (offshore)
• Automatic staff-culling via an awful yearly stack ranking reviews process (the dreaded PBC, or "Personal Business Commitment")
• Fading technical expertise and inability to retain top technical people
• Extreme bureaucracy and lack of agility
• Acquisitions focus instead of home-grown innovation
• Decline in key businesses
• Inability to be price competitive with the current leaders in Cloud

Ultimately, the company's lackluster performance is the responsibility of the chief executive, as well as her managment team. And that may mean that fresh leadership at the company may be required to turn things around.